I used to love an em-dash, the tripling effect and a provocative if/then sentence structure. Now I find myself working hard to remove these “tells” of AI writing, which have become synonymous with large language model (LLM) discourse. If gratuitous LinkedIn posts were irritating to you before, I’ve got bad news: it’s become worse. Now, every “I’m-delighted-to-announce-my-promotion” and “five-things-my-overseas-holiday-taught-me-about-business-deals” follows a formulaic writing style that has you questioning how Pat from HR who battled to string a sentence together a few months back is crafting Cannes-winning copy for his daily social media posts. Of course, the answer is ChatGPT. Or Claude. Or another AI tool that has helped elevate our language abilities ― to the detriment of trust and authenticity. These LLMs scour the internet in nanoseconds to pull patterns of successful writing styles and serve them up on a semantic plate for you to sample. The ease of “cut and paste” has become all too tempting. Without question, the rise of AI accessibility brings with it a range of benefits: saving time, enhancing productivity and boosting creativity. We can all enjoy instant, high-quality content generation, 24/7 Q&A support and multilingual capabilities. But it’s also created a monster, because now we don’t trust anything we read, by anyone. And, even more annoyingly as someone in the communications industry, I’m adjusting my own writing to sound “less Pat”. Behavioural linguistics has taught us that trustworthy, authentic brand communication leads to better client engagement, retention and loyalty. When our AI spidey sense is tingling and we don’t trust what we’re reading, we end up disconnecting from not only the message but the messenger. This erodes any credibility that has been meticulously built up over the years and it can be “painful and embarrassing” when AI is used covertly and discovered ― just ask our department of communication & digital technology. Brands and business leaders are aware of it. Increasingly, perfect prose and solid spelling hint towards AI-generated copy. And so, to help course-correct, we have entered an era of keyboard chaos: making mistakes ― on purpose ― to signal “human”. A new AI app called Sinceerly (yes, deliberately spelt that way) will add typos and grammatical errors into your content for a monthly subscription of $4.99. Initially created as a joke by a Harvard Business School graduate, the concept has taken off and you can now choose between three tiers of error ― “subtle”, “human” and “CEO”, to unlock a combination of typos, misspellings, incorrect use of lower case and the addition of (this one made me laugh) “sent from my iPhone”. All with a goal to make your emails sound less like they were written by ChatGPT. Apart from sounding more human, these new typo tools are also designed to help combat AI detectors, which are increasingly good at identifying AI text. Projected to hit more than $13bn by 2035, the AI detector market is booming. GPTZero, QuillBot, Originality.ai and ZeroGPT lead the way by measuring predictability, sentence variety and pattern recognition to identify stylistic fingerprints that tend to be unique to the different AI tools. But while typos might be the new communication status symbol right now, as we ― and our AI detector tools ― wise up patterns will undoubtedly change and the pendulum will swing. It’s a slippery slope. The question is: what is this all doing to our brains? We’re already seeing that too much reliance on LLMs can lead to mental decline, called “cognitive atrophy”. A recent study by the Massachusetts Institute of Technology (MIT) shows that consistent AI users struggle more with attention, memory and critical thinking due to weaker neural engagement. Here, brain imaging revealed a decline in cognitive engagement. Simply put, just like any muscle, when we don’t use parts of our brain it deteriorates. The researchers likened this type of cognitive debt to an overreliance on navigational tools. “Just as relying on a GPS dulls our sense of direction, relying on AI to write and reason can dull our ability to do those very things ourselves,” notes the MIT report. While this type of cognitive offloading isn’t new (after all, we’ve seen this film before with calculators and calendars), AI has taken it to a whole new level. “When your assistant can ‘think’ for you, you may stop thinking altogether,” is its cautionary conclusion. Whether you’re pro-AI or not, a fine balance seems to be the sensible approach. Perhaps the next tech innovation will be “cognitive wearables” to make sure you’re hitting your (semantic) steps. And maybe Vitality will throw in a smoothie. Futurists: you heard it here first. • Crymble is a behavioural linguist at BreadCrumbs Linguistics, a marketing firm that specialises in authentic behavioural communication. This article was entirely human written.